Wireless networks provide treacherous environments for digital communication. They are bandwidth-limited and error-prone. Their characteristics are unpredictable and time-varying. Transmitting interactive video over wireless mobile networks is even more challenging because of the bandwidth-hungry nature of video and the susceptibility of compressed video to data losses. Frequent data losses introduce an error (i.e., distortion) into displayed images which propagates to following images. The propagation is due to the temporal dependency of encoded video frames to their previously decoded images. The timing-varying nature of the networks further aggravates the problem as network characteristics (e.g., available bandwidth, latency, and loss rates) may undergo rapid and dramatic variations.
In this proposal, the project director proposes to study a new class of error recovery techniques that focuses on eliminating error propagation. Prior work has mainly taken an approach that prevents display errors from occurring from the first place. This existing approach is not effective for wireless interactive video transmission because data losses inevitably occur in wireless communication, and repairing them tends to introduce the delays in video playout. The approach is to isolate errors when they occur by preventing them from propagating. The delays in repairing data losses affect only the duration of error propagation. The project calls the approach, Recovery from Error Spread using Continuous Updates (RESCU). RESCU does not introduce any delay in video playout, and has potential to achieve good error resilience. A similar approach is adopted in several recovery techniques stipulated in H.263+, an International Telecommunication Union (ITU) low-bit rate standard. However, they are limited in that they require feedback channels and are not suitable for multicast. The proposed techniques, in contrast, have potential to work well with or without feedback channels, and to be scalable for multicast. Encouraging preliminary results indicate that such potential is highly realizable.
The research plan centers on making RESCU highly adaptive to timing-varying environments, and on evaluating its utility for interactive video transmission over wireless mobile networks. Four four major research components have been identified which comprehensively cover research issues arising from developing highly adaptive robust transmission and coding techniques.
* Adaptable video codecs.The first plan is to design adaptable coding schemes incorporating RESCU. The schemes are combined with transport-level recovery techniques to produce powerful recovery mechanisms which overcome the shortcomings of previous approaches.
* Adaptation. Given adaptable RESCU schemes with adjustable coding parameters, the research will tackle the problem of determining the values of these parameters appropriate for the maximum robustness under given network conditions. The approach is to apply rigorous Markovian modeling and analysis of the impact of these parameters to the final image quality and resource utilization.
* Transport-level services. Adaptable RESCU techniques require several transport-level functions: (1) accurate report on the current traffic characteristics; (2) efficient deadline-driven automatic repeat request (ARQ) techniques; and (3) flow and congestion control for robust interactive video communication. Ttechniques will be investigated to support these functions in the transport layer.
* Multicast support. Finally, the project will study the issues of robust video multicast including scalability for many receivers, receiver-driven control of adaptability, and fairness among heterogeneous receivers.
Education plan: The primary objective in education is to create a new generation of computer scientists who can accelerate the deployment of video technology into homes and workplaces. To realize the goal, the following activities are planned: (1) using videoconferencing as a primary means for instruction; (2) participating in multi-university initiative for deploying video technology in campus for instructional purpose; (3) extending opportunities for women, under-represented minorities, and undergraduate in research; and (4) developing a series of course dealing with video technology, multimedia, computer networking, and distributed systems.
